Neonatal and early‐onset diabetes in Ukraine: Atypical features and mortality

Abstract Aims The aim of this study is to elucidate the aetiology and clinical features of neonatal and early‐onset diabetes in a large database for pediatric diabetes patients in Ukraine. Methods We established a Pediatric Diabetes Register to identify patients diagnosed with diabetes before 9 months of age. Genetic testing was undertaken for 66 patients from 65 unrelated families with diabetes diagnosed within the first 6 months of life (neonatal diabetes, n = 36) or between 6 and 9 months (early‐onset diabetes, n = 30). Results We determined the genetic aetiology in 86.1% of patients (31/36) diagnosed before 6 months and in 20% (6/30) diagnosed between 6 and 9 months. Fourteen individuals (37.8% of those with a genetic cause identified) had activating heterozygous variants in ABCC8 or KCNJ11. An additional 10 individuals had pathogenic variants in the INS or GCK genes, while 4 had 6q24 transient neonatal diabetes. Rare genetic subtypes (including pathogenic variants in EIF2AK3, GLIS3, INSR, PDX1, LRBA, RFX6 and FOXP3) were identified in nine probands (24.3% of solved cases), 6 of whom died. In total, eight individuals died between infancy and childhood, all of them were diagnosed before 6 months and had received a genetic diagnosis. Conclusions In the last decade, the increased availability of comprehensive genetic testing has resulted in increased recognition of the contribution of rare genetic subtypes within pediatric diabetes cohorts. In our study, we identified a high mortality rate among these patients.

cause for their disease and are most likely to have type 1 diabetes. Pathogenic activating variants in the genes encoding the ATP-sensitive potassium channel (KATP) subunits (KCNJ11 and ABCC8), dominant variants in the INS gene and chromosome 6q24 methylation abnormalities are the most common causes of neonatal diabetes in European countries. [2][3][4][5] Pathogenic variants affecting over 25 additional genes cause rarer subtypes of the disease, 6 with 1-2 novel causes being reported each year.
The genetic causes of neonatal and early-onset diabetes vary depending on geography, ancestry and consanguinity. While cohort studies in Europe, Japan and the United States have reported de novo variants in KCNJ11 as the most common cause of permanent neonatal diabetes mellitus (PNDM), 4,[7][8][9][10][11][12] in countries with a high rate of consanguineous marriages, the inheritance pattern of PNDM is very different. In these settings, autosomal recessive subtypes are more common, with homozygous EIF2AK3 variants causing Wolcott-Rallison syndrome being the most common cause of PNDM. 13,14 Wolcott-Rallison syndrome is a syndromic form of neonatal/early-onset diabetes associated with a poor prognosis and high mortality rate in infancy. 15 Early mortality in patients with neonatal/earlyonset diabetes has been reported in individuals with other rare genetic subtypes, such as those with monogenic autoimmune diabetes (including IPEX syndrome and diabetes caused by recessive LRBA variants) 16,17 and Donohue syndrome. 18 The contribution of rare genetic subtypes, including those discovered in the last 5 years, in European countries is not clear, as is the disease's mortality rate.
We previously reported the genetic causes in 42 cases of neonatal and early-onset diabetes occurring within the first 9 months of life in Ukraine and investigated treatment change in patients with KCNJ11 or ABCC8 pathogenic variants. 19 Following on from this study, we now report 24 additional patients and focus our investigation on rare genetic causes diagnosed in recent years, atypical features and mortality rate.

| Subjects
A neonatal and early-onset diabetes section of the Ukrainian Pediatric Diabetes Registry (UPDR) was created in 2012 to include individuals diagnosed with diabetes before 9 months of age identified by regional Ukrainian pediatric endocrinologists. Annual meetings on reconciliation of UPDR with regional endocrinologists were held by the Ukrainian authors of the article in accordance with the order of the Ministry of Health of Ukraine. All patients received insulin therapy free of charge according to the UPDR, so all patients with type 1 diabetes were included and the coverage was 100%. The Registry also contains data on patients with type 2 and monogenic diabetes. 20 We present data from the UPDR starting from Jan 2013 to Dec 2021.
GAD and IA2 antibody testing was performed in individuals diagnosed between 6 and 9 months. The 30 individuals who were negative for both were further investigated through genetic testing.

| Genetic testing
Genetic testing was undertaken using a combination of Sanger sequencing, targeted next-generation-sequencing (tNGS) for all known neonatal diabetes genes (full list available on https://www.diabe tesge nes.org) and methylation analysis for chromosome 6q24 abnormalities as previously described. 13 A type 1 diabetes (T1D) genetic

"What is already known?"
Neonatal diabetes mellitus is defined as diabetes diagnosed in the first 6 months of life, with some cases presenting between 6 and 9 months of age (early-onset diabetes). While cohort studies in Europe, Japan and the United States have reported de novo variants in KCNJ11 as the most common cause of permanent neonatal diabetes mellitus (PNDM), in countries with high rate of consanguineous marriages the inheritance pattern of PNDM is very different, with autosomal recessive causes being most common.
"What this study has found?" Our data show a high proportion of rare genetic causes, including autosomal recessive aetiologies, within our cohort (24.4%), and a high mortality rate (21.6% of individuals with a confirmed genetic diagnosis). To our knowledge, this is the first study assessing mortality in a large neonatal/ early-onset diabetes cohort in a European country. Clinical follow-up in our cohort highlighted atypical presentation and clinical course of diabetes in some families.
"What are the implications of the study?" Implementation of early comprehensive genetic testing including targeted next-generation sequencing can improve clinical management of this genetically heterogeneous disease. risk score (T1D GRS) was calculated by next-generation sequencing of 30 T1D-associated SNPs. 21 Patient 38 was tested by the Invitae laboratory (USA) using their tNGS Monogenic Diabetes Panel.

| Clinical features
Clinical features at diagnosis and subsequent follow-up were collected from all patients. For 42 individuals who were previously reported, we collected follow-up information from hospital records and follow-up visits.

| Statistical analysis
Clinical characteristics are presented as median (interquartile range). For quantitative data Kruskal-Wallis test was used for comparative statistics.

| RESULTS
We identified 70 cases with diabetes diagnosed before 9 months of age from the UPDR. DNA samples were available for 66 individuals from 65 unrelated families (94.2%). When available, variant testing was performed in family members by Sanger sequencing (primers available on request). The initial presentation for 42 of these individuals has been previously described. 19 Thirty-six individuals within our cohort were diagnosed before 6 months and therefore had neonatal diabetes. Diabetes remitted in 10 of them (27.8%), resulting in a diagnosis of transient neonatal diabetes. Thirty individuals had antibody-negative, early-onset diabetes diagnosed between 6 and 9 months. Summary of clinical characteristics for the two groups are presented in Table 1.
Disease-causing variants were identified in 37 patients. This included 31/36 (86.1%) individuals diagnosed before 6 months (neonatal diabetes) and 6/30 (20%) individuals diagnosed between 6 and 9 months (early-onset diabetes) ( Figure 1 and Table 2). Fourteen probands (37.8% of solved cases) harbored pathogenic activating variants in one of the KATP channel genes, ABCC8 or KCNJ11. Dominant variants in the INS and GCK genes were identified in six and four individuals, respectively. Methylation abnormalities at 6q24 were identified in four individuals. Causative variants in the rarer causative genes (EIF2AK3, GLIS3, INSR, PDX1, LRBA, RFX6 and FOXP3) were found in a total of nine individuals (24.4% of solved cases), six of whom (66.7%) died in early infancy/childhood.
In total 8 individuals in our cohort died between the ages of 3 months and 9 years. All of them were diagnosed T A B L E 1 Clinical characteristics of diabetes in 0-6 and 6-9 months groups before the age of 6 months and had a genetic diagnosis. The causes of death included renal insufficiency (1), cerebral edema (n = 1) and pneumonia (n = 1). The cause of death was not known in five cases.

Rare genetic subtypes (n = 9 patients)
EIF2AK3 gene (n = 3 patients, 2 previously reported at presentation 19  FOXP3 gene (n = 1 patient) Patient 33 was born from a second pregnancy on the background of retrochorial hematoma. Polyhydramnios, signs of intrauterine infection and fetal distress were recorded at 32 weeks of gestation. The male proband was born at 40-41 weeks gestation with a birth weight of 3880 g and an Apgar score of 6/7 points. Since birth, the child presented with severe respiratory disorders, with nasal constant positive airway pressure needed from the seventh day of life. At 5 days, the child was transferred to the Regional Children's Clinical Hospital to the neonatal intensive care unit. The child's condition on admission was extremely severe due to diaphragmatic hernia, respiratory failure and suppressed reflexes. On the 16th day of life, a right thoracotomy was performed to correct the right diaphragmatic hernia. After surgery, the child was on mechanical ventilation. An attempt to switch to spontaneous breathing at 23 days was unsuccessful due to respiratory failure. On the 30th day of life, he was diagnosed with diabetes (BGL up to 23.8 mmol/L), and insulin therapy was started. Glycemic control was challenging and satisfactory compensation of diabetes could not be achieved. At 32 days, his C-peptide was 0.21 ng/ml (normal value 0.9-7.1). General blood tests, biochemical analysis and urine tests did not show significant changes. A diffuse mucopurulent endobronchitis was detected by a bronchoscopy. The child died aged 90 days. Three months after his death, genetic testing identified a pathogenic, maternally inherited hemizygous FOXP3 variant, p.(Arg347His).

LRBA gene (n = 1 patient)
The proband (Patient 34) was born at 38 weeks gestation with a weight of 3150 g. He was the second child of unrelated healthy parents (an elder sibling was unaffected, a second and third pregnancy had previously terminated with their miscarriage). He was first admitted to the infectious disease department of a district hospital at the age of 5 months with a suspicion of acute intestinal infection. He was treated with antibacterial therapy and parenteral rehydration without significant effect. Two mg of dexamethasone was administered parenterally four times (a total of 8 mg) per day. The child was referred to the NICU where infusion therapy and antibiotics were continued. Further investigations revealed hyperglycemia and decreased C-peptide levels (<0.05 ng/ml, normal value 0.81-3.85).
Insulin therapy was started. The child's condition initially improved, and the boy was discharged in a satisfactory condition to continue outpatient treatment with insulin. Genetic testing was performed in 2014, but no pathogenic variants were identified in the 20 genes known to cause neonatal diabetes at that time. He was admitted again at the age of 9 months, with a preliminary diagnosis of acute gastroenterocolitis, disseminated intravascular coagulation syndrome (DICS), intestinal toxicosis and exicosis (grade 2). Due to further deterioration of his condition, with suspicion of ulcerative necrotic enterocolitis, ileal gangrene, serous peritonitis and intestinal obstruction, an urgent diagnostic laparoscopy was performed, which was extended to direct laparotomy. Given the significant pathological changes in the intestine due to necrotic ulcerative colitis, mesenteric thrombosis, intestinal necrosis, and serous hemorrhagic peritonitis, resection of 40 cm of the ileum and ileostomy were performed. Subsequently, the patient's condition did not improve, and he died of cardiac arrest. The post-mortem found an abdominal form of nodular periarteritis, complicated by total hemorrhagic gangrene of the small and large intestine, diffuse seroushemorrhagic peritonitis as well as focal hemorrhages in the lungs. Four years later, following the identification of LRBA biallelic variants as a cause of neonatal/early-onset diabetes, 16 further testing was performed and a homozygous frameshift variant in the LRBA gene p.(Glu946Ter) was identified.

Neonatal and early-onset diabetes due to the common genetic causes (n = 24 patients, 18 previously reported at presentation 19 )
Activating pathogenic variants in ABCC8 or KCNJ11 were identified in 14 patients from 13 unrelated families diagnosed before 6 months (8 with KCNJ11 and 5 with ABCC8), and one individual diagnosed between 6 and 9 months (who was heterozygous for a KCNJ11 pathogenic variant). Four individuals with ABCC8 variants and one with a heterozygous KCNJ11 variant had TNDM with remission between 1 and 12 months after presentation. The patient with KCNJ11-TNDM (Patient 8 in Table 2) was diagnosed with juvenile rheumatoid arthritis at 8 years and received insulin temporarily during treatment with glucocorticoids. The heterozygous pathogenic KCNJ11 and ABCC8 variants had arisen de novo in 9 cases (2/5 ABCC8 and 7/9 KCNJ11). One patient (Patient 12 in Table 2) was compound heterozygous for the ABCC8 variants. 19 The remaining four cases had inherited the pathogenic variant from their mothers, two of whom (both harboring KCNJ11 variants) were diagnosed with diabetes at 3 months of age and had been treated with insulin until the genetic diagnosis in their children (aged 29 and 35), (see Table 2). Transfer to sulfonylurea therapy was successful in all probands with activating KCNJ11 and ABCC8 variants in our cohort 19 including the affected mothers. Three patients with ABCC8 pathogenic variants had atypical clinical features at follow-up. Patient 12 who was compound heterozygote for the ABCC8 variants p.(Val-324Met)/p.(Arg1394Leu) had severe neurological features which did not improve when sulfonylurea (SU) treatment was started when he was 6 years old. At the age of 8, he presented with growth failure (height − 4.5 SD) and inguinal cryptorchidism. Additional examinations detected hypopituitarism with low IGF-1 level (81 ng/ml, normal range 95-460). Treatment with chorionic gonadotropin showed a temporary effect. An orchidopexy was performed when he was 11. Treatment with growth hormone was started at 13 and has not led to worsening of the glycemic control or increasing of SU dose. At the age of 13, the patient still has severe generalized hypotonia and is unable to sit, hold his head upright, walk or talk. Two siblings heterozygous for a de novo ABCC8 p.(Ile49Phe) variant (Patients 10 and 11 in Table 2) had severe developmental delay, epilepsy, and neonatal diabetes (DEND) syndrome. The proband was diagnosed with diabetes at 3 months which remitted at 1 year and relapsed at 2 years. His sister developed convulsions and hypoglycemic coma at 5 months, and at 6 months was diagnosed with diabetes and received insulin for a few days before being transferred to SU treatment. Both siblings had been treated with a low SU dose with excellent glycemic control (HbA 1c was stable <7%) until their death. The proband died aged 9 years, 1 day after admission to the ECU because of hyperthermia, cytolysis syndrome, jaundice and systemic multiple organ failure of unknown origin. His 5-year-old sister died of pneumonia (she also had severe rickets and curvature of the chest).
The second most common genetic cause of neonatal/ early-onset diabetes in our cohort were heterozygous pathogenic variants in the INS gene (n = 6 patients, three diagnosed before 6 months). In one family, the pathogenic INS p.(Gly32Ser) variant was also detected in the patient's mother and maternal grandmother who were diagnosed with 'type 1 diabetes' at 3 years. The disease presentation was atypical in the proband who had neonatal hypoglycemia on the 3rd day of life (glucose level was 2.1 mmol/L), and despite developing diabetes at 5 months, did not start insulin treatment until 2 years of age. We observed a similar presentation in another individual with the same INS p.(Gly32Ser) variant who did not need insulin treatment between 2 and 8 months of age. 19 Thus, two patients with heterozygous INS variants in our cohort had remission of diabetes.
Transient neonatal diabetes caused by 6q24 paternal uniparental disomy was identified in 4 patients (3 previously reported 19 ). In two cases the diabetes relapsed at 9 and 10 years (HbA 1c -8.0% and 8.5%, respectively); both are currently treated with diet only. The other two cases are currently aged 2 and 8 years and are still in the remission phase.
Heterozygous GCK variants were identified in 4 individuals (Patients 28-31 in Table 2, one previously reported 19,22 ). Two of them were diagnosed before the age of 6 months. In all 4 cases, BGL was checked in the asymptomatic infants as their mothers had been diagnosed with gestational diabetes which had not required insulin treatment. Despite the early presentation, all patients underwent genetic testing after 1 year of age, and their fasting hyperglycemia was between 6.1 and 6.9 mmol/L. Patient 30 had one hypoglycemic episode 2 hours after birth. In family 29, the mother and proband's brother, who were diagnosed with diabetes at 30 and 18 years, respectively, were also heterozygous for the pathogenic GCK variant, confirming a diagnosis of GCK-MODY.

Neonatal diabetes diagnosed before 6 months without a confirmed genetic diagnosis (n = 5)
No likely causative variant was identified in five patients diagnosed before 6 months of age (Table 3). Patient 38 was diagnosed with impaired glucose tolerance (IGT) at 5 months and 3 weeks and was asymptomatic. She had moderately elevated BGL, low C-peptide 0.4 ng/ml (normal range 0.81-3.85), and negative pancreatic autoantibodies. Analysis of all the known monogenic diabetes genes did not identify a likely cause; however, she was noted to be heterozygous for a HNF1A p.(Thr537Arg) variant of uncertain significance. Further investigations are needed to determine whether the variant is linked to the phenotype or if it is a benign polymorphism. Patient 39 likely had hyperglycemia of prematurity since he was born at 26.5 weeks and the diabetes remitted when he was 4.5 months of age. The remaining 3 unsolved neonatal diabetes patients all had a high T1D GRS suggesting that they are likely to have very early-onset type 1 diabetes.

| DISCUSSION
In this study we assessed the aetiologies and clinical features of neonatal and early-onset diabetes occurring within the first 9 months of life in 66 children from Ukraine. Our data show a high proportion of rare genetic causes within our cohort (24.4% of solved cases), including autosomal recessive aetiologies, and a high mortality rate (21.6% of individuals with a confirmed genetic diagnosis). To our knowledge, this is the first study assessing mortality in a large neonatal and early-onset diabetes cohort in a European country. Over the last decade, the identification of novel, rarer causes of neonatal/early-onset diabetes and the availability of comprehensive testing has allowed us to increase our diagnostic yield with almost all our cases with diabetes diagnosed before 6 months having a likely explanation for their early presentation (86.1% having a monogenic cause identified, 8.4% having early-onset type 1 diabetes, 2.8% having hyperglycemia of prematurity). Testing in individuals diagnosed between 6 and 9 months who were antibody negative identified a genetic cause in 20% of cases, highlighting the importance of considering genetic testing also in this disease group where T1D is much more common.
Our comprehensive testing approach has allowed us to identify cases with rare genetic aetiologies and cases with atypical presentation. Despite Ukraine being a country with a low rate of consanguineous unions, autosomal recessive causes of neonatal/early-onset diabetes were identified in eight families. In two of these families the parents were known to be related, with compound heterozygous variants identified in the remaining six cases. Our results show the important contribution of autosomal recessive causes of neonatal/early-onset diabetes, even in a nonconsanguineous setting.
An early genetic diagnosis has important implications for the patients, allowing targeted management of their diabetes. An example of this in our cohort was the early detection of individuals heterozygous for GCK pathogenic variants, who are often misdiagnosed. We identified four children with heterozygous GCK variants, consistent with them having fasting hyperglycemia from birth (GCK-MODY). GCK-MODY is more commonly diagnosed in adults where hyperglycemia is incidentally picked up; however, in some cases it can be detected in the neonatal period, 19,22 most commonly due to control of BGL in children by their affected mothers. A genetic diagnosis of GCK-MODY is important in these patients to avoid unnecessary treatment and monitoring and to inform management of pregnancy in affected mothers.
Clinical follow-up in our cohort highlighted atypical presentation and clinical course of diabetes in some families. Three of our patients (Patient 11 with a heterozygous ABCC8 variant, patient 16 with a maternally inherited INS variant, and patient 30 with a heterozygous GCK variant) presented with hypoglycemia in the neonatal period (patient 11 presented with hypoglycemic coma). Coexistence of hypoglycemias and hyperglycemias has been reported in patients with loss-of-function ABCC8 hyperinsulinism, 23 resulting from dysregulation of insulin secretion in patients with diffuse ABCC8 hyperinsulinism. 24 We can hypothesize that similar dysregulation of insulin secretion may occur also in patients with gain-of-function mutations in the KATP channel genes; however, such a presentation in those with pathogenic INS and GCK gene variants, to the best of our knowledge, is novel. While in our cases we cannot exclude completely the possibility of transitional hypoglycemia, this atypical presentation requires further investigation. Furthermore, we observed remission of diabetes in two patients with heterozygous pathogenic INS variants (in one case for more than 1 year). This is unusual as dominant INS variants most commonly cause permanent diabetes.
Some of the additional extra-pancreatic features identified in our cohort were also atypical, for example the presence of hypopituitarism in patient 12 who is compound heterozygous for two ABCC8 variants. This case also had severe neurological features which did not improve with sulphonylurea treatment, this could be due to delayed diagnosis and treatment with sulphonylurea. [25][26][27][28] Furthermore, patient 36, who was homozygous for a PDX1 start-loss variant, did not have symptoms of pancreatitis and/or malabsorption, but had intrahepatic biliary tract hypoplasia and died in infancy. This is different from the phenotype of isolated neonatal diabetes with or without exocrine insufficiency described in previous reports of individuals with recessive PDX1 pathogenic variants. 29,30 Patient 33, who was hemizygous for a pathogenic FOXP3 variant, also had an atypical clinical course for IPEX syndrome (Immune dysregulation, Polyendocrinopathy, Enteropathy, X-linked), and did not have any of the classical features of this condition 31 other than diabetes, although it is possible that these would have developed if the child had survived. For all patients in our cohort, we only performed genetic testing for the known genetic causes of neonatal/early-onset diabetes, we, therefore, cannot exclude the presence of other genetic variants, which may contribute to the extra-pancreatic features observed.
For some of the individuals with the rarer genetic subtypes, the extra-pancreatic features were consistent with the genetic diagnosis. For example, Mitchell-Riley syndrome caused by biallelic RFX6 pathogenic variants is characterized by neonatal/early-onset diabetes, pancreatic hypoplasia, intestinal atresia, and gallbladder aplasia or hypoplasia. Early mortality has been reported in some patients. 32 Patient 37 in our cohort had typical features of this syndrome, with the exception of the presence of erosive-hemorrhagic proctosigmoiditis. Similarly, patient 34 had a classical presentation of LRBA-related disease, 16 namely diabetes and enteropathy. Timely haematopoietic stem cell transplantation 33 or use of abatacept and/or glucocorticoid therapy can result in a more favorable prognosis in individuals with LRBA-related disease.
We observed a relatively high mortality rate within our cohort, with eight patients dying between infancy and childhood (21.6% of cases with a confirmed diagnosis). All these patients were diagnosed with diabetes before 6 months of age and had a genetic cause identified. The mortality rate is even higher when only considering patients with rare genetic subtypes (6/9, 66.7%). The higher mortality in this group is likely to be due to many different factors, including the fact that extra-pancreatic complications are common in these patients and therapeutic options, which improve prognosis are currently available only in those with some monogenic forms of autoimmune diabetes. 16,17,33 Our study had some limitations. First, ZnT8 antibodies were not routinely tested in individuals diagnosed between 6 and 9 months, which may have negatively affected the pick-up rate in this age group. Furthermore, since we used tNGS for the known genetic causes of diabetes, we cannot exclude the possibility of additional variants in non-diabetes genes contributing to the atypical presentation and mortality in some cases.

| CONCLUSIONS
The present study highlights the broad spectrum of genetic heterogeneity and clinical presentations of neonatal and early-onset diabetes in Ukraine, where rare autosomal and X-linked recessive genetic subtypes, atypical findings and high mortality rate were relatively common. Whereas autoimmune monogenic diabetes, Donohue syndrome and Wolcott-Rallison syndrome are known to have a poor prognosis, death in infancy/childhood in patients with PDX1 and ABCC8 pathogenic variants is rare. Implementation of early comprehensive genetic testing including targeted next-generation sequencing can improve clinical management of this genetically heterogeneous disease.

AUTHOR CONTRIBUTION
Evgenia Globa performed a clinical investigation of patients at the initial stage and follow-up, was responsible for conception and design of the study, data acquisition, preparation of the manuscript, finding relevant references, and final approval of the manuscript. Nataliya Zelinska performed a clinical investigation of patients at the initial stage and follow-up; designed the analyses; reviewed and edited the manuscript. Matthew B Johnson, Sarah E. Flanagan and Elisa De Franco performed and interpreted genetic testing; conceptualized and designed the study; and critically reviewed and revised the manuscript. Elisa De Franco is the guarantor and approved the final manuscript as submitted.